Madhan Karthikeyan

AI Researcher

Chennai, Tamil Nadu, India0 mo experience

AI EnabledAI ML Practitioner

Key Highlights

Developed advanced AI solutions for tech support automation.
Achieved high accuracy in food recognition using computer vision.
Strong foundation in AI/ML with hands-on project experience.

Stackforce AI infers this person is a skilled AI/ML intern with practical experience in computer vision and intelligent systems.

Contact

Skills

Core Skills

Computer VisionArtificial Intelligence (ai)

Other Skills

Image SegmentationYOLOv8D-FINE modelsDataset curationModel evaluationIntelligent AgentsSemantic SearchOCRFastAPIPostgreSQLVector embeddingsCommunicationTeam ManagementRedisCelery

About

VIT University | Pre-final Year | B.Tech in Computer Science IIT Madras | B.S. in Data Science (Ongoing) Interested in AI and Machine Learning Focused on building practical skills through projects and hands-on learning Open to internships and opportunities in AI/ML

Experience

0 mo

Total Experience

Average Tenure

Current Experience

Globally gi

Computer Vision Intern

Feb 2026 – Present · 4 mos · Chennai, Tamil Nadu, India · Hybrid

Developed a computer vision pipeline for multi-object food recognition using YOLOv8 and D-FINE models. Curated and annotated a 5,000-image dataset spanning 50 food categories and implemented class-balancing and targeted augmentation strategies to improve model robustness. Achieved 80–85% mAP across food classes while reducing per-class performance variance by approximately 35% on challenging occluded and multi-object scenes.

Computer VisionImage Segmentation

National informatics centre, govt of india

AI Intern

Jun 2025 – Jul 2025 · 1 mo · Chennai, Tamil Nadu, India · On-site

AI Intern – Support Management System | July 2025
As part of my AI internship, I undertook the development of an AI-powered Support Management System aimed at automating tech support tasks. The system features two core AI components:
1) Screenshot to Form Autofill – Leveraged OCR (Tesseract, EasyOCR), regex, and vision-language models like Qwen2.5-VL (7B) to extract error details directly from user-submitted screenshots and autofill tech support forms, significantly reducing manual effort.
2) AI-Driven Solution Retrieval – Built a RAG (Retrieval-Augmented Generation) pipeline using FastAPI and PostgreSQL with vector embeddings to retrieve accurate resolutions from official documentation based on the submitted issue.
Key tools explored included OpenAI, Gemini APIs, and local models (LLaVA, Moondream). The end-to-end system automated the entire support flow—from visual issue detection to intelligent solution recommendation.
Impact: Reduced manual response time significantly, increased support efficiency, and laid the foundation for a fully automated AI-based helpdesk system.

Intelligent AgentsSemantic SearchArtificial Intelligence (AI)